Development of a Forecasting Model for Dengue Infection in Health Region 8 Using SARIMA and Holt-Winters Exponential Smoothing: Python

Abstract

Dengue infection is a significant public health issue in Health Region 8. This study aims to develop and evaluate the performance of monthly dengue case forecasting models using Seasonal Autoregressive Integrated Moving Average (SARIMA) and Holt-Winters Exponential Smoothing techniques to support disease control planning. The study utilized dengue case data from Health Region 8 during 2020–2024. The dataset was divided into a training set (2020–2023) and a testing set (2024). The SARIMA model was constructed by analyzing stationarity and determining parameters from The Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plots. The additive Holt-Winters Exponential Smoothing method was chosen to accommodate data exhibiting trends and seasonality. Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE). The results showed that The SARIMA (2,0,0) (0,0,0)[12] model with intercept yielded MAE = 213.96, RMSE = 330.00, and MAPE = 49.65%, effectively forecasting the overall trend but lacking precision in detecting outbreak peaks. In contrast, the Holt-Winters Exponential Smoothing (Additive) model achieved MAE = 193.02, RMSE = 234.03, and MAPE = 52.32%, reflecting seasonal patterns more closely to the observed data. For the conclusion and recommendation, The Holt-Winters model demonstrated superior absolute accuracy (MAE and RMSE) compared to SARIMA in monthly dengue case forecasting, particularly in capturing seasonal patterns, although it exhibited a slightly higher MAPE. Both models offer distinct advantages: SARIMA is suitable for long-term policy planning, while Holt-Winters is more appropriate for short-term preparedness and response. Further research integrating additional factors is recommended to enhance forecasting accuracy.